Search

Precise and prompt predictions of crop yields are crucial for optimising farm management, post-harvest operations, and marketing strategies within the agricultural sector. While various machine learning approaches have been employed to forecast crop yield, their application to grain quality, particularly head rice yield (HRY), is less explored. This research collated crop-level HRY data across four seasons (2017/18–2020/21) from Australia's rice-growing region. Models were developed using the XGBoost algorithm trained at varying time steps up to 16 weeks pre-harvest. The study compared the accuracy of models trained on datasets with climate data alone or paired with vegetative indices using two- and four-week aggregations. The results suggest that model accuracy increases as the harvest date approaches. The dataset combining climate and vegetative indices aggregated over two weeks surpassed industry benchmarks early in the season, achieving the highest accuracy two weeks before harvest (LCCC = 0.65; RMSE = 6.43). The analysis revealed that HRY correlates strongly with agroclimatic conditions nearer harvest, with the significance of vegetative indices-based features increasing as the season progresses. These features, indicative of crop and grain maturity, could aid growers in determining optimal harvest timing. This investigation offers valuable insights into grain quality forecasting, presenting a model adaptable to other regions with accessible climate and satellite data, consequently enhancing farm- and industry-level decision-making. [ABSTRACT FROM AUTHOR]

Wheat is the important food grain and is cultivated in many Indian states: Punjab, Haryana, Uttar Pradesh, and Madhya Pradesh, which contributes to major crop production in India. In this study, popular statistical approach multiple linear regression (MLR) and time series approaches Time Delay Neural Network (TDNN) and ARIMAX models were envisaged for wheat yield forecast using weather parameters for a case study area, i.e., Junagarh district, western Gujarat region situated at the foot of Mount Girnar. Weather data corresponds to 19 weeks (42nd to 8th Standard Meteorological Week, SMW) during crop growing season was used for prediction of wheat yield using these statistical techniques and were evaluated for their predictive capability. Furthermore, trend analysis among weather parameters and crop yield was also carried out in this study using non-parametric Mann–Kendall test and Sen's slope method. Significant negative correlation was observed between wheat yield and some of the weekly weather variables, viz., maximum temperature (48, 49, 50, 51, 52, and 4th SMW), and total rainfall (50, 51, and 1st SMW) while positive correlation was observed with morning relative humidity (49 and 3rd SMW). Study indicated that forecast error varied from 1.80 to 10.28 in MLR, 0.79 to 7.79 in ARIMAX (2,2,2), − 3.09 to 10.18 in TDNN (4,5) during model training period (1985–2014). The MAPE value shows that the time series data predicted less than 5% of variation, whereas the conventional MLR technique indicated more than 7% variation. Both ARIMAX and TDNN approaches indicated better performance during model training periods, i.e., 1985–2014 and 1985–2015, while former performed well during the forecast periods 1985–2016 and 1985–2017. Overall, the study indicated that the ARIMAX approach can be used consistently for 4 years using the same model. [ABSTRACT FROM AUTHOR]

In recent years, Deep Learning Multi-Layer Perceptron (DLMLP) neural networks have shown remarkable success in addressing crop yield forecast related problems. The methodologies used so far for crop yield forecast with remotely sensed data were focused upon vegetation indices generated from optical data. The prediction of crop yield in an accurate manner by developing robust machine learning models based on soil health parameters is crucial since it helps keep a track of soil health as well as its impact on overall yield. This study aims to utilize remotely sensed Microwave satellite data from Sentinel-1 and optical data from Sentinel-2, and field data to estimate three important soil health parameters- Soil Moisture, Soil Salinity, and Soil Organic Carbon (SOC). The study has been carried out in the Rupnagar district of Punjab in India. The estimated soil health parameters, SAR backscatter, and optical remote sensing satellite data parameters were utilized to estimate wheat crop yield. The soil health based DLMLP model performed best in crop yield estimation and gave R2 values of 0.723 and 0.684 in the training and testing phases, respectively, and Mean Absolute Error (MAE) of 0.98 and Root Mean Square Error (RMSE) value of 1.24 for the 2019–20 season. The DLMLP test R2 was 42.2% more than the Ordinary Least Squares Regressor (OLS), while the MAE and RMSE were 37.97% and 38.61% less than the OLS regressor for wheat crop yield estimation. The soil health-based DLMLP model gave satisfactory yield estimation accuracy in the absence of validation of soil health parameter values for the preceding years-2015–16 till 2018–19 wheat seasons. This study's novel feature is that it estimates soil health parameters for the early stages of wheat crop growth when soil lies mostly exposed and utilises them for crop yield prediction.

Accurate forecast of corn yields is important for decision making regarding food and energy management strategies. In this work, we developed an unprecedented optimized framework for the MODIS-based mid-season corn yield forecasting over five producing states of the United States: Illinois, Indiana, Iowa, Nebraska, and Ohio. We evaluated Enhanced Vegetation Index (EVI)-based forecasts under different schemes accounting for: different machine learning techniques with a mid-season composite or multi-temporal composites as inputs, four (county-, district-, state-, and global-based) training domains, and 16-day composites versus daily interpolated composites involving the day of pixels as predictors. Under the best identified scheme, we compared the EVI-based forecasts with those based on Normalized Difference Vegetation Index (NDVI), Leaf area Index (LAI) and Fraction of Absorbed Photosynthetic Active Radiation (FPAR). EVI and NDVI were transformed to LAI (named as LAIEVI and LAINDVI) as predictors for producing EVI- and NDVI-based forecasts. We evaluated both county- and state-level forecasts using the percent error (PE), mean absolute PE (MAPE) and determination coefficient (R2). The linear regression models driven by the single latest composite in mid-season often outperformed elastic net and random forest models driven by multi-temporal composites. The forecast performance decreased with longer subsets of EVI composites being used. The performance under the different training domains varied by states and forecast level (county or state), although the changes within states were mostly non-significant except in Nebraska. The forecasts based on 16-day and daily composites performed similarly, indicating that the use of information about the day of pixel composite provides no additional benefit to the yield forecast. For the best EVI-based schemes, the medium annual MAPE (PE) at the county (state) level varied between 6.1% (2.4%) and 7.7% (5.3%) across states while the medium annual R2 (interannual R2) varied between 0.54 (0.59) and 0.82 (0.86). Results suggested that, while EVI was, in general, the best predictor for the Corn Belt as a whole, the adequacy of the EVI- and NDVI-based forecasts varied by states and largely exceeded that of the LAI- and FPAR-based forecasts. Compared with the EVI-based forecasts, the NDVI-based forecasts performed better in Iowa (MAPE’s 0.9% and 1.43% lower at the county and state level), similar in Nebraska and worse in the other states. Overall, the best state-level forecasts consistently outperformed concurrent National Agricultural Statistical Service (NASS) forecasts.

Accurate crop yield forecasting can help stakeholders take effective measures in advance to avoid potential grain supply risks. However, currently, yield forecasts are mostly made close to harvest (e.g. 1–3 months before harvest for Chinese winter wheat), which gives stakeholders a relatively short time to react, decide, and intervene. To satisfy stakeholders' requirements for timely and precise yield forecasting, we propose a hybrid machine learning-enabled early-season yield forecasting method integrated with an intermediate climate forecast process. The results show that: (1) Compared with the baseline model, our proposed method advances winter wheat yield prediction up to 8 months before harvest with satisfactory accuracy. (2) The climate forecast process incorporated is effective and consistently optimized in various model combinations and controlled experiments. (3) The proposed method performs robustly over different spatial scales (e.g., in the first month of Chinese winter wheat, the yield predictive accuracy is improved in 183 out of 233 counties). In summary, our work provides an effective and robust approach for early-season yield forecasting that gives stakeholders more time to take appropriate actions to cope with crop yield volatility risks. • Timely forecasting of crop yields is critical to regional and global food supplies. • An early-season yield forecast framework with hybrid machine learning is proposed. • Our proposed methodology outperforms the benchmark models in the early season. • Our methodology exhibits strong robustness across heterogeneous regions in China. [ABSTRACT FROM AUTHOR]

• Wheat and maize yield anomalies for the Pannonian basin are forecasted using XGBoost. • Maize yield anomalies can be forecasted accurately two months before harvest. • Impact of severe droughts on crop yield losses remains underestimated. • Forecasting of temporal yield variability is more reliable than spatial variability. • Soil moisture is the most important predictor for maize yields in drought years. The increasing frequency and intensity of severe droughts over recent decades have led to substantial crop yield losses in the Pannonian Basin in southeastern Europe. Their socioeconomic consequences can be minimized by accurate crop yield forecasts, but such forecasts often underestimate the impact of severe droughts on crop yields. We developed a gradient-boosting-based crop yield anomaly forecasting system for the Pannonian Basin and examined its performance, with a focus on drought years. Winter wheat and maize yield anomalies are forecasted for 42 regions in the Pannonian Basin using predictor datasets from Earth observation and reanalysis describing vegetation state, weather, and soil moisture conditions. Our results show that crop yield anomaly estimates in the two months preceding harvest have better performance (maize errors 14–17%, wheat 13–14%) than earlier in the year (maize errors 21%, wheat 17%). The forecast models can satisfactorily capture the interannual yield anomalies, but spatial yield variability is only partially reproduced. In years of severe drought, the wheat model performs better than under average conditions with errors below 12%. The errors of the maize forecasts in drought years are larger than average forecast skill: 31% two months ahead and 20% one month ahead. However, for both crops the yield losses remain underestimated by the forecasts in severe drought years. The feature importance analysis shows that during the last two months before harvest, wheat yield anomalies are controlled by temperature and evaporation and maize by the combined effects of temperature and water availability as expressed by several drought indices. In severe drought years, during the two months before harvest the seasonal temperature forecast becomes the most important predictor for the wheat forecasts and soil moisture for the maize model. Overall, this study provides in-depth insights into the impact of droughts on crop yield forecasts in the Pannonian Basin. [ABSTRACT FROM AUTHOR]

Olive trees are one of the most economically important perennial crops in Portugal. During the last decade, the Alentejo olive-growing region has suffered a significantly change in the crop production system, with the regional pollen index (RPI) and olive fruit production registering a significant growth. The aim of this study was to ascertain the utility of this highly variable production and pollen data in crop forecasting modeling. Airborne pollen was sampled using a Cour-type trap from 1999 to 2015. A linear regression model fitted with the regional pollen index as the independent variable showed an accuracy of 87% in estimating olives fruit production in Alentejo. However, the average deviation between observed and modeled production was 32% with half of the tested years presenting deviations between 36 and 66%. The low accuracy of this model is a consequence of the great overall variation and significant upward trend observed in both the production and the RPI dataset that conceal the true association between these variables. In order to overcome this problem, a detrend procedure was applied to both time series to remove the trend observed. The regression model fitted with the fruit production and the RPI detrended data showed a lowest forecasting accuracy of 63% but the average deviation between observed and modeled production decrease to 14% with a maximum deviation value of 33%. This procedure allows focusing the analysis on the production fluctuations related to the biological response of the trees rather than with the changes in the production system. [ABSTRACT FROM AUTHOR]

The long-term forecast of yield and vegetation conditions of grain crops for the dry-steppe zone of the pre-Ural Region is made for the first time based on new predictors and their time lags. The forecast is developed for the value of the predictant ( Y) and all predictors ( X) that are included in the multiple regression models. The obtained results permit the probability of the onset of drought in the expected year to be described in advance (5-6 months before the start of field works). The developed models indicate that a strong drought is expected for grain crops in 2017 (the winter wheat yield of 2-3 c grain/ha), which is due to unfavorable weather conditions in June (the precipitation of 3-4 mm) and high air temperature in July (the average temperature in the first decade of 25.2°C and maximum temperature of 37.5°C). This information makes it possible to choose the optimal technological solutions for overcoming the influence of unfavorable conditions. The methods of step-by-step multiple regression and neural network regression and the method of residual deviations in combination with the method of superposition of epochs are used. [ABSTRACT FROM AUTHOR]

The objective of this work was to evaluate an estimation system for rice yield in Brazil, based on simple agrometeorological models and on the technological level of production systems. This estimation system incorporates the conceptual basis proposed by Doorenbos & Kassam for potential and attainable yields with empirical adjusts for maximum yield and crop sensitivity to water deficit, considering five categories of rice yield. Rice yield was estimated from 2000/2001 to 2007/2008, and compared to IBGE yield data. Regression analyses between model estimates and data from IBGE surveys resulted in significant coefficients of determination, with less dispersion in the South than in the North and Northeast regions of the country. Index of model efficiency (E1') ranged from 0.01 in the lower yield classes to 0.45 in higher ones, and mean absolute error ranged from 58 to 250 kg ha‑1, respectively.O objetivo deste trabalho foi avaliar um sistema de estimativa de produtividade da cultura do arroz no Brasil, com base em modelos agrometeorológicos simples e no nível tecnológico dos sistemas produtivos. Este sistema de estimativa incorpora as bases conceituais propostas por Doorenbos & Kassam para produtividades potenciais e alcançáveis com ajustes empíricos para produtividade máxima e sensibilidade da cultura à deficiência hídrica, considerando-se cinco classes de produtividade de arroz. A produtividade de arroz foi estimada de 2000/2001 a 2007/2008 e comparada aos dados do IBGE. Análises de regressão entre as estimativas do modelo e dados do IBGE resultaram em coeficientes de determinação significativos, com menor dispersão na região Sul do que nas regiões Norte e Nordeste do país. O índice de eficiência do modelo (E1') variou de 0,01, nas classes de menor produtividade, a 0,45 nas classes de maior produtividade, e o erro médio absoluto variou de 58 a 250 kg ha‑1, respectivamente.

• A crop yield and soil moisture database is established for the Upper Blue Nile Basin, Ethiopia. • An easy-to-implement analog-based model is developed for real-time crop yield forecast. • The analog-based model is verified to have considerable forecast skills in various environments. As Ethiopia's population grows, crop yield predictions are becoming increasingly important for national food security. Accurate, easy-to-implement, and computationally efficient forecast approaches are desirable for broad applications in emerging regimes like Ethiopia. In this study, we develop and test an analog approach for pre-season crop yield prediction conditioned on antecedent precipitation and planting time soil moisture content indices to guide cultivation decision making. Historical planting time soil moisture at four selected sites were simulated using the Coupled Routing and Excess STorage (CREST) hydrological model and classified into five levels. Likewise, a historical crop yield database for each of the five classes of planting time soil moisture were constructed using the Decision Support System for Agrotechnology Transfer (DSSAT) agricultural model. Adopting maize as a representative crop, analog models based on different indexes or predictors with various lead times were constructed and used to conduct hindcasts during 1979–2014 and real-time forecast in 2018 and 2019. Both the hindcast and real-time forecasts were then evaluated against yield observations. To verify the applicability at locations with various environments, the analog models were then applied in different Agroecological Zones. The analog models were shown to be accurate and easy to implement, which may incentivize adoption by local extension agents and regional agricultural agencies to inform farmers' crop choices. [ABSTRACT FROM AUTHOR]

A wine forecast model for one of the most arid wine regions of the Europe—Alentejo was improved and tested for the period 1998–2014. During this period, Alentejo region had strong upward trends in wine production associated to the increase of vineyard area. The forecast model was supported on a hierarchical analysis, including the determination of the potential production at flowering by quantifying airborne pollen concentration, followed by a climate based evaluation of the possible impact of fruit-set conditions in the limitation of production. Through the monitoring of airborne pollen flows it is possible to define an accurate main pollen season and determine the regional pollen index that will be used as independent variable in the regional forecast model. The time trend, which was initially removed from data, was then added back to obtain the forecast. Stepwise regression and cross-validation were employed during the period 1998–2014 for calibration of the model used for predicting annual wine production. The developed model explained about 86% of wine variance over the years with absolute average error of 6% for the cross validation and 87% of cases had differences between actual and forecasted wine production below 10%. The reliability and early-indication ability of the proposed forecast model justify their use to respond to a number of government agencies and wine industry concerns and activities. [ABSTRACT FROM AUTHOR]

Airborne pollen emission model was used to determine pollen metrics and to examine their relationship with vineyard phenology in two wine regions of Northern Portugal: Vinhos Verdes (1993-2007) and Douro (1992-2011). A number of airborne pollen metrics were obtained through the rate of changes of logistic model adjusted to the time series of airborne pollen. In both regions, the mean absolute differences between observed phenology and model-predicted values for start, peak and final of flowering phenophases were always lower than 5 days and the slope of the regression through the origin is close to one. These metrics can be used to accurately and precisely predict the dynamic of Vitis flowering observed at field level. The model's simplicity and flexibility are of great advantage for its practical use in aerobiology. [ABSTRACT FROM AUTHOR]

Sea surface temperatures (SSTs) can provide valuable predictive information for regional climate and agricultural yields in many parts of the world. This study was conducted to identify relationships between Atlantic and Pacific SSTs and corn yields in Alabama, Florida, and Georgia and to use these relationships to develop forecasts that can be used at lead times prior to spring planting. The relationships between seasonal SSTs and detrended county corn yields were analysed using singular value decomposition (SVD) analysis and confirmed using principal component analysis (PCA). Using a Monte Carlo approach, field-significant results were found with SSTs in the July-September (JAS) and October-December (OND) seasons in the previous year and with the January-March (JFM) season of the current. A significant portion of the variability of SSTs with corn yield residuals was connected with anomalous heating and cooling associated with the El Niño Southern Oscillation (ENSO). Based on the results found by SVD analysis and confirmed by PCA, indices of spatially averaged SSTs in regions of the North Pacific and Atlantic Oceans were derived. Using these indices along with the Niño3.4 index to represent tropical Pacific SSTs, cross-validated multiple linear regression models were developed to predict corn yield residuals using index values in the JAS and OND seasons. Using the cross-validated models 91.5 and 98.4% of forecasted county corn yield residuals showed predictive skill (based on tercile hit scores) with seasonal index values in the JAS and OND seasons, respectively. The results of the models indicate that the indices of SSTs show significant predictability with corn yield residuals at lead times up to 4-7 months prior to spring planting and are a significant improvement over the use of an index of ENSO alone. Copyright © 2010 Royal Meteorological Society [ABSTRACT FROM AUTHOR]

The article discusses the evaluation of the impact of large-scale oceanic and atmospheric climate models on corn yields in the U.S. Correlation analysis was used to determine the link between detrended corn yield residuals, precipitation, and surface temperature as well as climate indices. It notes that significant associations were determined with indices of the Pacific-North American model, tropical North Atlantic, and eastern tropical Pacific sea surface temperatures.

Abstract: Global/Regional Circulation Models (GCM/RCM) predict the interannual climate variability better than the absolute values of meteorological variables. Statistical bias-correction methods increase the quality of daily model predictions of incoming solar radiation, maximum and minimum temperatures and rainfall frequency and amount. However, when bias-corrected forecasts/hindcasts are used by dynamic crop models, timing of dry-spell occurrences generate the largest uncertainty during the linking process. In this study, we used 20 ensemble members of an 18-year period provided by the Florida State University/Center for Ocean-Atmospheric Prediction Studies (FSU/COAPS) regional spectral model coupled to the National Center for Atmospheric Research Community Land Model (CLM2). The daily seasonal-climate hindcast was bias-corrected and used as input to the CERES-Maize model, thus producing 20 crop yield ensemble members. Using observed weather data for the same period, a time series of simulated crop yields was produced. Finally, principal component (PC) regression analysis was used to predict this time series using the crop yield ensemble members as predictors. Between 13.7 and 28.8% of the simulated corn yield interannual variability was explained using only one principal component (p <0.05), and estimated yields were in the correct tercile by margins of 16.7 to 38.2% beyond chance. Predictability of simulated corn yields using principal components was improved relative to the use of bias-corrected daily hindcasts. Bias-correcting all meteorological variables used by the crop model increased predictability skills compared with use of raw hindcasts, individual bias-correction of rainfall, and climatological values. [Copyright &y& Elsevier]

Three existing models are coupled to assess crop development and forecast yield in the largest contiguous irrigation network in the world: the Indus Basin in Pakistan. Monteith’s model is used for the calculation of absorbed photosynthetically active radiation (APAR), the Carnegie Institution Stanford model is used for determining the light use efficiency, and the surface energy balance algorithm for land (SEBAL) is used to describe the spatio-temporal variability in land wetness conditions. The new model requires a crop identification map and some standard meteorological measurements as inputs. The conversion of above ground dry biomass into crop yield has been calibrated through harvest indices and the values obtained are compared with the international literature. The computations were executed in a GIS environment using 20 satellite measurements of the advanced very high resolution radiometer (AVHRR) to cover an annual crop rotation cycle. The validation with district data revealed a root mean square error of 525, 616, 551 and 13,484 kg ha−1 for wheat, rice, cotton and sugarcane yield, respectively. The model performs satisfactorily for wheat, rice and sugarcane, and poorly for cotton. It is expected that the accuracy of the model applied to 1.1 km pixels decreases with the increasing number of crops occurring within a given pixel. Although AVHRR is basically too coarse a resolution for field scale crop yield estimations, the results provides yield predictions to policy makers in Pakistan with a spatial detail that is better than the traditional district level data. The gaps between the average and the maximum yield are 1075 and 1246 kg ha−1 for wheat and rice, respectively. Future work should rely on the integration of high and low resolution images to estimate field scale crop yields. [Copyright &y& Elsevier]

• Develop a within-growing season yield forecast system with random forest model. • Random forest model performs well in predicting grain yield in China. • We identified the most important stage-specific predictors determining crop yield. • The most important variable influencing yields varied with crop types. Accurate and timely crop yield forecasts can provide essential information to make conclusive agricultural policies and to conduct investments. Recent studies have used different machine learning techniques to develop such yield forecast systems for single crops at regional scales. However, no study has used multiple sources of environmental predictors (climate, soil, and vegetation) to forecast yields for three major crops in China. In this study, we adopted 7-year observed crop yield data (2013–2019) for three major grain crops (wheat, maize, and rice) across China, and three major data sets including climate, vegetation indices, and soil properties were used to develop a dynamic yield forecasting system based on the random forest (RF) model. The RF model showed good performance for estimating yields of all three crops with correlation coefficient (r) higher than 0.75 and normalized root means square errors (nRMSE) lower than 18.0%. Our results also showed that crop yields can be satisfactorily forecasted at one to three months prior to harvest. The optimum lead time for yield forecasting depended on crop types. In addition, we found the major predictors influencing crop yield varied between crops. In general, solar radiation and vegetation indices (especially during jointing to milk development stages) were identified as the main predictor for winter wheat; vegetation indices (throughout the growing season) and drought (especially during emergence to tasseling stages) were the most important predictors for spring maize; soil moisture (throughout the growing season) was the dominant predictor for summer maize, late rice, and mid rice; precipitation (especially during booting to heading stages) was the main predictor for early rice. Our study provides insights into practical crop yield forecasting and the understanding of yield response to environmental conditions at a large scale across China. The methods undertaken in this research can be easily implemented in other countries with available information on climate, soil, and vegetation conditions. [ABSTRACT FROM AUTHOR]